Keyword Error Detection and Correction on Vietnamese Product Data Using Symmetric Deletion Algorithm

Unlike normal text data that is standardized in terms of words and strict grammar, Vietnamese e-commerce product data is a special type of data created by system users, including many keywords related to brands and trademarks. In many cases, sellers aim to advertise or highlight text on e-commerce platforms, ignoring the grammar and keyword order in a text sentence. This paper proposes an effective method based on the symmetric deletion algorithm suitable for large and noisy data sets to identify misspelled keywords and recommend their original keywords. In the experiment, the proposed method is used to integrate into a system working on a real e-commerce data set, with two main functions: (1) searching for product text, to do this we use the Elastic Search tool with input text processed through the proposed method; (2) identifying misspelled keywords and giving their original keywords. The quality assessment part is done through a real labeled dataset.